arXiv:astro-ph/9710106v1 10 Oct 1997 Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 1 February 2008 (MN L A T E X style file v1.4) The radio–optical correlation in steep-spectrum quasars Stephen Serjeant 1,2 , Steve Rawlings 2 , Stephen J. Maddox 3 , Joanne C. Baker 4 , Dave Clements 5 , Mark Lacy 2 , Per B. Lilje 6 1 Astrophysics Dept., Blackett Labs., Imperial College London, Prince Consort Road, London SW7 2BZ 2 Astrophysics, Department of Physics, University of Oxford, Keble Road, Oxford OX1 3RH 3 Royal Greenwich Observatory, Madingley Road, Cambridge, CB3 0EZ 4 Mullard Radio Astronomy Observatory, Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE. 5 European Southern Observatory, Karl-Schwarzschild-Strasse 2, D-85748 Garching-bei-Munchen, Germany. 6 Institute of Theoretical Astrophysics, University of Oslo, P.O. Box 1029 Blindern, N-0315 Oslo, Norway 1 February 2008 ABSTRACT Using complete samples of steep-spectrum quasars, we present evidence for a correlation between radio and optical luminosity which is not caused by selection effects, nor caused by an orientation dependence (such as relativistic beaming), nor a byproduct of cosmic evolution. We argue that this rules out models of jet formation in which there are no parameters in common with the production of the optical continuum. This is arguably the most direct evidence to date for a close link between accretion onto a black hole and the fuelling of relativistic jets. The correlation also provides a natural explanation for the presence of aligned optical/radio structures in only the most radio luminous high-redshift galaxies. 1 INTRODUCTION It is now established that quasars can be divided into two physically distinct classes: radio-loud quasars (RLQs) and radio-quiet quasars (RQQs) (Peacock, Miller & Longair 1986; Kellermann et al. 1989; Miller, Peacock & Mead 1991; Miller, Rawlings & Saunders 1993; Wilson & Colbert 1995). Relativistic jets are a universal feature of RLQs (e.g. Bridle et al. 1994) and are also probably associated with at least some RQQs (e.g. Miller et al. 1993); the difference between the classes is thus not whether the quasars can form relativistic jets, but related instead to the fraction of the total power output channelled along them in a bulk kinetic form (see also Rawlings 1994). It is now also established that at least some ra- dio galaxies harbour obscured quasar nuclei (e.g. An- tonucci 1993; Antonucci, Hurt & Kinney 1994; Dey & Spinrad 1996, Ogle et al. 1997). However, the popular notion that the probability of obscuration is a strong function of the angle between the jet axis and the line-of-sight (e.g. Barthel 1989; Antonucci 1993) due to the ostensible presence of a dusty molecular torus, is still debated, and may apply only to a restricted range of radio luminosity and/or redshift (see e.g. Lawrence 1991; Jackson & Rawlings 1997 and refs. therein). Accepting this orientation-based unification scheme, however, it is possible to combine RLQs and at least some radio galaxies into a single ‘radio-loud’ category, and estimate the power in the photoionising ultraviolet continuum Q phot (which is hidden in the case of radio galaxies) from the luminosity of the nar- row emission lines. By doing this in the 3CRR sample (Laing, Riley & Longair 1983) Rawlings & Saunders (1991) inferred that ‘radio-loud’ objects have bulk powers Q bulk in their jets of the same order as Q phot , whereas in the case of RQQs (and their possible ob- scured counterparts; see Lonsdale, Smith & Lonsdale 1989) the fraction of the power output channeled into jets is > ∼ 10 3 times lower (Miller et al. 1993; Rawl- ings 1994). The strongly differing fractions seem to be a fundamental difference between radio-loud and radio-quiet active galactic nuclei. Despite this difference there is some evidence that within each class of quasar there is some intrinsic connection between Q bulk and Q phot . If one accepts that radio luminosity (L rad ) and [O iii] line luminos- ity (L [O III] ) are crude indicators of these variables then the separate, but roughly parallel, loci of the radio-loud and radio-quiet objects in the L rad versus c 0000 RAS
Transcript
arX
iv:a
stro
-ph/
9710
106v
1 1
0 O
ct 1
997
Mon. Not. R. Astron. Soc. 000, 000–000 (0000) Printed 1 February 2008 (MN LATEX style file v1.4)
The radio–optical correlation in steep-spectrum
quasars
Stephen Serjeant1,2, Steve Rawlings2, Stephen J. Maddox3,
Joanne C. Baker4, Dave Clements5, Mark Lacy2, Per B. Lilje6
1Astrophysics Dept., Blackett Labs., Imperial College London, Prince Consort Road, London SW7 2BZ2Astrophysics, Department of Physics, University of Oxford, Keble Road, Oxford OX1 3RH3Royal Greenwich Observatory, Madingley Road, Cambridge, CB3 0EZ4Mullard Radio Astronomy Observatory, Cavendish Laboratory, Madingley Road, Cambridge CB3 0HE.5European Southern Observatory, Karl-Schwarzschild-Strasse 2, D-85748 Garching-bei-Munchen, Germany.6Institute of Theoretical Astrophysics, University of Oslo, P.O. Box 1029 Blindern, N-0315 Oslo, Norway
1 February 2008
ABSTRACT
Using complete samples of steep-spectrum quasars, we present evidencefor a correlation between radio and optical luminosity which is not causedby selection effects, nor caused by an orientation dependence (such asrelativistic beaming), nor a byproduct of cosmic evolution. We argue thatthis rules out models of jet formation in which there are no parameters incommon with the production of the optical continuum. This is arguablythe most direct evidence to date for a close link between accretion onto ablack hole and the fuelling of relativistic jets. The correlation also providesa natural explanation for the presence of aligned optical/radio structuresin only the most radio luminous high-redshift galaxies.
1 INTRODUCTION
It is now established that quasars can be dividedinto two physically distinct classes: radio-loud quasars(RLQs) and radio-quiet quasars (RQQs) (Peacock,Miller & Longair 1986; Kellermann et al. 1989; Miller,Peacock & Mead 1991; Miller, Rawlings & Saunders1993; Wilson & Colbert 1995). Relativistic jets area universal feature of RLQs (e.g. Bridle et al. 1994)and are also probably associated with at least someRQQs (e.g. Miller et al. 1993); the difference betweenthe classes is thus not whether the quasars can formrelativistic jets, but related instead to the fraction ofthe total power output channelled along them in abulk kinetic form (see also Rawlings 1994).
It is now also established that at least some ra-dio galaxies harbour obscured quasar nuclei (e.g. An-tonucci 1993; Antonucci, Hurt & Kinney 1994; Dey &Spinrad 1996, Ogle et al. 1997). However, the popularnotion that the probability of obscuration is a strongfunction of the angle between the jet axis and theline-of-sight (e.g. Barthel 1989; Antonucci 1993) dueto the ostensible presence of a dusty molecular torus,is still debated, and may apply only to a restrictedrange of radio luminosity and/or redshift (see e.g.
Lawrence 1991; Jackson & Rawlings 1997 and refs.therein). Accepting this orientation-based unificationscheme, however, it is possible to combine RLQs andat least some radio galaxies into a single ‘radio-loud’category, and estimate the power in the photoionisingultraviolet continuum Qphot (which is hidden in thecase of radio galaxies) from the luminosity of the nar-row emission lines. By doing this in the 3CRR sample(Laing, Riley & Longair 1983) Rawlings & Saunders(1991) inferred that ‘radio-loud’ objects have bulkpowers Qbulk in their jets of the same order as Qphot,whereas in the case of RQQs (and their possible ob-scured counterparts; see Lonsdale, Smith & Lonsdale1989) the fraction of the power output channeled intojets is >
∼ 103 times lower (Miller et al. 1993; Rawl-ings 1994). The strongly differing fractions seem tobe a fundamental difference between radio-loud andradio-quiet active galactic nuclei.
Despite this difference there is some evidencethat within each class of quasar there is some intrinsicconnection between Qbulk and Qphot. If one acceptsthat radio luminosity (Lrad) and [O iii] line luminos-ity (L[O III]) are crude indicators of these variablesthen the separate, but roughly parallel, loci of theradio-loud and radio-quiet objects in the Lrad versus
L[O III] plane (e.g. Rawlings 1994) hints at a correla-tion between Qbulk and Qphot whether the jet is ina radio-loud quasar (with Qbulk ∼ Qphot) or a radio-quiet quasar (with Qbulk ≪ Qphot). This gives rise tothe hope that these relations might constrain modelsfor the fuelling of relativistic jets in all types of activegalaxies (e.g. Rawlings & Saunders 1991; Falcke et al.
1995).
However, this is far from being a universally ac-cepted view. In both radio-loud and radio-quiet ob-jects, the correlation between Qbulk and the inferredQphot can be influenced, or even caused by selectioneffects in samples used to date, as can the apparentlysimilar magnitudes of Qbulk and Qphot in radio-loudobjects. We will discuss this in more detail in section3.4.
There are further objections, on physical ratherthan statistical grounds. For example, Dunlop & Pea-cock (1993) have suggested that Lrad–L[O III] corre-lations in radio-loud objects are more plausibly ex-plained by models in which both luminosities are en-hanced through interactions between radio jets (andlobes) and a dense environment (see also Baum, Zirbel& O’Dea 1995 and refs. therein). There are certainlysome cases of both radio-loud objects (e.g. Lacy &Rawlings 1994) and radio-quiet objects (e.g. Axon et
al. 1989) where at least some of L[O III] may be at-tributable to power supplied by the jet. A key ques-tion, therefore, is whether one can find a direct linkbetween Qbulk and Qphot rather than inferring thelatter indirectly from L[O III].
The optical continuua and nuclear emission linesof RLQs are the direct probes of Qphot required. Dun-lop & Peacock (1993) cite the lack of a correlation be-tween the optical continuum and radio luminosities ofRLQs as an argument in favour of their environmentalinterpretation of the Lrad–L[O III] correlation. In factthe existing evidence on the radio–optical correlationfor RLQs is mixed (Peacock, Miller & Longair 1986;Browne & Murphy 1987; Neff, Hutchings & Gower1989; Miller, Peacock & Mead 1990; Miller et al. 1993- we will review these contradictory claims in section3.4) and, in practice, confused by a number of issues.
Firstly, Peacock, Miller & Longair (1986) showedthat one must not mix RLQs and RQQs since do-ing so can produce an apparently universal (but spu-rious) radio–optical correlation. The authors arguedthat RLQs have a minimum optical luminosity to ex-plain why 3C quasars are among the brightest op-tically, and why the RLQs in the optically-selectedBright Quasar Survey (Schmidt & Green 1983) areamong the most luminous radio sources. This alonewould yield a spurious correlation when comparingRLQs with RQQs over a wide optical range. (In factour new data rules out this scenario, as we discuss insection 3.4.)
Secondly, the strong orientation dependence ofthe optical and radio continuua in flat-spectrum core-dominated RLQs (e.g. Jackson et al. 1989; Baker
1996), can lead to a radio–optical correlation. Thisfollows because radio and optical luminosities can besimultaneously enhanced in these objects by relativis-tic beaming of synchrotron radiation.
Thirdly, although there are hints of cou-pled optical–X-ray–radio processes in ostensibly un-beamed RLQs (Browne & Wright 1985; Browne &Murphy 1987), the samples used to date either lackthe spectroscopic redshifts which would distinguishradio luminosity dependence from evolution, or haveserious and unquantifiable selection biases.
In this paper we present the results of a newstudy of the radio–optical correlation for quasars inwhich we have attempted to implement lessons learntfrom the studies referenced above. We confined ourattention to one of the distinct classes of quasars, theRLQs, since we believe their radio luminosities are amore straightforward indicator of Qbulk (which, withsuitable radio data, it is possible to estimate in in-dividual quasars; Rawlings & Saunders 1991; Rawl-ings 1993). A recent study of the radio–optical cor-relation in RQQs can be found in Lonsdale et al.
1995. To reduce the effects of relativistic beamingand optical continuum anisotropy we have also cho-sen to concentrate on steep-spectrum RLQs (here-after SSQs), to the exclusion of flat-spectrum, core-dominated quasars. Section 2 outlines the selectionand study of our new complete sample of SSQs. Todistinguish between Lrad-dependence and evolution(z dependence), this sample is combined with twoother complete radio-selected samples which togetherspan a wide range in Lrad at a given redshift; we alsomake a comparison with the SSQs in an optically-selected sample. In Section 3 we present evidence fora radio–optical correlation, and assess the role of sam-ple selection. In Section 4 we attempt a physical in-terpretation of the radio–optical correlation, and com-ment briefly on the implications of this result for stud-ies of radio galaxies. Throughout this paper, we takevalues of H0 = 100 kms−1Mpc−1, Ω = 1 and a zerocosmological constant.
2 MOLONGLO-APM QUASAR SURVEY
Our new, complete sample of SSQs was drawn ini-tially from the 408-MHz Molonglo Reference Cata-logue (MRC, Large et al. 1981), down to a flux den-sity S408 ≥ 0.95 Jy with a radio spectral index cut-off α ≥ 0.5 (where α = −d log Sν/d log ν evalu-ated near 1 GHz). The MAQS survey area is lim-ited to a ∼ 1-sr region in which both UK SchmidtAPM (Automated Plate-measuring Machine) dataand Texas Catalogue positions (Douglas et al. 1996)were available, bounded roughly by −35 <
∼ δ <∼
0, 21h <∼ α <
∼ 5h. Hence, the sample is calledthe Molonglo/APM Quasar Survey (MAQS; Serjeant1996, Maddox et al. in preparation, Serjeant et al.
in preparation). Optical identifications were made
Figure 1: The radio luminosity (L408), redshift (z) plane for SSQs from the MAQS (squares), the MQS(stars), 3CRR (circles) and the BQS (crosses). The brightest 33% (excluding the BQS) have been plottedas open symbols. For 3CRR the 178MHz-750MHz spectral indices from Laing, Riley & Longair (1983) wereused. The loci of sources with α = 0.85 and S408 = 0.95 (MAQS/MQS limit) and 4.94 Jy (3CRR limit) areshown as dashed lines.
from the UK Schmidt bJ plates, using APM classifica-tions to exclude “galaxy” identifications brighter thanbJ = 20.5. The quasar candidates therefore consistedof all “stellar” or “merger” identifications, and all“galaxy” identifications fainter than bJ = 20.5. Red-shifts have been measured (at the Anglo-Australian,William Herschel and Nordic Optical Telescopes) forall quasar candidates to the plate limit of bJ ≃ 22.5,yielding a complete spectroscopic catalogue of 159(steep- and flat-spectrum) confirmed quasars. Opti-cal spectra of all quasar candidates (including thoserejected for lacking broad optical-UV lines) will ap-pear in a forthcoming paper.
The deep plate limit is very important to thisstudy: only then is the survey likely to detect essen-tially all the SSQs as far as z ≃ 3.3 where the Lyαline leaves the bJ band. A recent spectroscopic studyof a complete and radio-fainter sample of 7C radiosources (Willott et al. 1997) quantifies this by findingno SSQs optically fainter than the plate limit usedfor the MAQS (i.e. SSQs with bJ > 22.5 compriseless than 5 per cent of the total population), but thatmost are fainter than the B ≈ 20 limit of studiesbased, for example, on POSS-I.
We have combined data on our new sample with
data on SSQs from the radio-selected 3CRR sample(Laing, Riley & Longair 1983). In addition, we aug-ment the MAQS sample with SSQs taken from theMolonglo Quasar Sample (MQS, Baker 1994), also se-lected from the MRC down to S408 = 0.95 Jy, and tothe bJ plate limit. Finally, we include the SSQs fromthe optically selected Bright Quasar Survey (hereafterBQS, Schmidt & Green 1983). Figure 1 shows 408-MHz radio luminosity as a function of redshift (L408,z) for the four samples; the MQS points are plottedin separate symbols where the identification is notshared by the MAQS. Taken together, these four sam-ples provide a wide dispersion in L408 at any redshiftz, allowing us to distinguish cosmic evolution (z de-pendence) from radio luminosity dependence.
We have split the combined sample into two fluxdensity bins; open symbols are used for the 33% ofsources with S408 > 3.23 Jy, and filled symbols for theremainder. This split was chosen to yield a roughlyeven fraction in the brighter bin throughout the red-shift range. Measured values of radio spectral index, αwere used where available (100% of 3CRR and BQS,97% of MQS, 72% of MAQS), and α = 0.85 assumedotherwise. Some of the brighter SSQs, and nearly allthe flat spectrum quasars, should appear in the Parkes
Figure 3: The variation of apparent magnitude withredshift for a quasar of fixed absolute magnitude. Thetwo curves show the variation for a numerical integra-tion of the composite quasar spectrum, and the cor-responding variation with a power law optical quasarcontinuum (as discussed in the text). The power lawmodel is the brighter at a redshift of z = 5. Both mod-els have been normalised to produce the same apparentmagnitude at a redshift of 1.5.
catalogue (Wright et al. 1991): for a source on theMAQS radio flux limit the Parkes catalogue shoulddetect all the radio sources with α <
∼ 0.4. We are thusconfident that flat-spectrum sources have been almostentirely excluded, despite our currently incomplete αinformation.
3 RESULTS
3.1 The radio–optical correlation
In Figure 2, optical luminosity is shown as a functionof z, again split into bright and faint radio subsam-ples. In order to calculate the absolute B-magnitude,MB, we transformed bJ magnitudes to B magnitudesusing B ≈ bJ − 0.14 and adopted typical optical spec-tral indices of 0.5 (the value derived e.g. from the com-posite 7C SSQ spectrum by Willott et al. 1997). Amore careful treatment of K-corrections was not pos-sible since a significant fraction of the objects are fromthe literature, having redshifts but not spectropho-tometry. Typical uncertainties in apparent magni-tudes are ∼ ±0.2 mag; similar contributions to theabsolute photometric errors are made by the uncer-tainties in the K-corrections. (Error bars are omittedfrom the plots for clarity.)
Two possible further sources of error are galacticreddening and emission line contributions to the bJ
flux (intrinsic reddening will be discussed separatelybelow). The galactic reddening is typically AB ≃ 0.2magnitudes, and the variations in galactic reddeningare expected to cause an additional relative photo-
metric error of ∼ 0.1 mags (de Vaucoleurs & Buta1983a, b). This variation is smaller than our typicalphotometric errors, so we do not correct for it here.The MAQS extends to galactic latitudes only as lowas | b |= 30, so any galactic reddening is likely to bemore significant for 3CRR; this would only increasethe significance of our results discussed below. Unfor-tunately not all the quasars from the literature havepublished line fluxes, so it is not possible to apply em-pirical emission line corrections to the bJ fluxes uni-formly over the sample. However, it is well-establishedin radio quiet quasar surveys that emission lines makea small contribution to the overall photometric errors,e.g. Schmidt & Green (1983). Since the bJ passbandis probably wider than the B selection passband ofSchmidt & Green (1983), we might expect the lineeffects to be even smaller. This is verified in Figure 3,where we have integrated the composite quasar spec-trum of Francis et al. (1991), appropriately redshifted,over the bJ passband. Also plotted is the predictionfor an αopt = 0.5 optical power law, which is the con-tinuum slope determined in the composite SSQ spec-trum of Willott et al. (1997). The deviations fromthe power-law model in Figure 3 are dominated bythe slightly atypical (for SSQs) optical spectral indexof the Francis et al. (1991) spectrum, except at red-shifts z >
∼ 3.3 where Lyα exits the bJ passband. Notethe smoothness of the composite quasar curve.
The loci of SSQs at three apparent magnitudesare shown as dashed lines in Figure 2; the upper linerepresents the magnitude limit of the BQS (B = 16),the middle line, the bJ = 20.5 limit above whichoptically-extended objects were removed from theMAQS, and the lower line the bJ ≃ 22.5 plate limit ofthe MAQS and MQS. From Figure 2, any incomplete-ness in the MAQS caused by rejecting bJ < 20.5 opti-cally extended objects is only expected to be presentat z < 0.4. Broad-line radio galaxies from 3CRR arealso optically extended on POSS plates, and have notbeen included. A point-like reddened 3CRR quasar,3C22, is plotted in Figure 2 as two asterisks joined bya line, the lower symbol representing the measuredMB and the upper representing the magnitude cor-rected for reddening (see Rawlings et al. 1995).
A clear tendency is seen for the radio-brightquasars at any z to also be brighter optically. Thistrend cannot be explained solely on the basis of selec-tion effects since, for example, the 3CRR SSQs haveno optical magnitude limit yet are brighter on thewhole than the MAQS/MQS quasars.
Comparing on Figure 2 the SSQs of MAQS/MQSwith those of 3CRR (restricting both to 0.4 < z < 2to ensure well matched comparisons), we find a meanapparent magnitude of 18.36 ± 0.20 for sources above3.23 Jy (open symbols), but 19.22 ± 0.14 for the re-mainder; the null hypothesis of identical distribu-tions is rejected at ≃ 99% confidence by both theKolmogorov-Smirnov and Mann-Whitney tests. Forthe combined samples in the restricted redshift range
Figure 2: The optical luminosity (MB), redshift (z) plane (symbols as Figure 1). The upper dashed linemarks the BQS magnitude limit, the lower line the MAQS limit, and the central line indicates the approx-imate threshold of reliable star-galaxy separation. The reddened quasar 3C22 is marked as two asterisksjoined by a vertical line (see text).
0.4 < z < 2, we find the optical and radio fluxes aresignificantly correlated (≫ 99.9% confidence usingSpearman’s ρ); the same is true of the whole sam-ple (i.e. unrestricted z, Figure 4). This is the firstevidence that the apparent radio–optical correlationis not dominated by a redshift effect.
Figure 5 shows the L408-MB correlation implicitin Figure 2. The correlation in Figure 5 is significantat ≫ 99.9% confidence (using Spearman’s ρ), with abest-fit slope of −2.5 × (0.6 ± 0.1) (in the log10 L408
vs. MB plane) and dispersion of ∼ 1.6 optical mag-nitudes (using 3CRR and MAQS/MQS). However,this slope and dispersion may be biased estimators ofthe true values, because the density of points on theplane depends on the weighting imposed by the bi-variate (radio–optical) luminosity function (LF), andthe sampled comoving volume.
Thus, by comparing the 3CRR and MAQS/MQSsamples we find evidence for the radio-optical cor-relation which (for the first time) is independent ofcosmic evolution. Is the correlation present in any ofthe samples separately? It might be argued that if acorrelation only appears when samples are combined,then it is more suspect than a correlation from a sam-ple in which selection effects are uniform across the
sources and where fewer corrections need to be madeto inter-compare the sources. In fact the correlationalso present in the MAQS in isolation (to a similar sig-nificance level), but it is perhaps instructive to showcircumstances in which this type of argument can bemisleading. Suppose we had two samples of quasars:the first has z = 1± 0.1 and B = 18± 0.1; the secondhas z = 1±0.1 and B = 22±0.1. It would probably beimpossible to demonstrate a radio-optical correlationwith either sample in isolation, but once combinedthe correlation would be far easier to detect.
The L408-MB correlation in Figure 5 is supportedby the radio properties of the optically faintest SSQs(the band between the lower two dotted lines in Fig-ure 2). With the exception of reddened quasars like3C22, which appears well above this band once cor-rected for reddening (Rawlings et al. 1995), this re-gion includes no 3CRR SSQs. It is, however, wellpopulated with MAQS/MQS SSQs. Although red-dened examples certainly exist (Baker & Hunstead1995, Willott et al. 1997), SSQs are typically red-dened much less severely than 3C22 (or are reddenedso severely, AV ≫ 1, that they will be classified asradio galaxies even on the basis of K-band images),since 3C22 has no clear broad emission lines in the
Figure 4: The flux-flux radio–optical relation for SSQs (symbols as Figure 1). Note the horizontal offsetbetween the filled (lower) and open (upper) symbols.
observer-frame optical. This argues strongly againstL408-dependent reddening as a cause of the radio–optical relation, an explanation which would in anycase require post-shock temperatures finely tuned tothe destruction of dust in, and only in, the most lu-minous radio sources.
This is not to say that reddening does not af-fect the correlation in any way; we only claim thatreddening alone cannot cause it. For example, it re-mains possible that reddening may affect the scatterin the correlation (see Section 4.2), or that the degreeof reddening may be linked to the optical luminositye.g. via dust sublimation.
Finally, although our samples have sufficient cov-erage of the radio-optical-redshift parameter spaceto demonstrate a radio-optical correlation, they areprobably not large enough to address any possibleevolution in this correlation. Interestingly, taking thehigher redshifts in isolation the evidence for the cor-relation is more marginal. However, this may simplybe due to lack of data, and the restricted dynamicrange in radio power at high redshifts⋆ (see Figure 1).The same high redshift radio-optical behaviour is also
⋆ Such a selection effect may also be responsible for the
seen in numerical simulations (discussed below) sup-porting this interpretation. Nevertheless, it remainspossible that the correlation weakens or perhaps failsat high redshifts, implying a different mechanism forjet formation at high (z >
∼ 2) and low (z ≃ 1) redshifts(section 4).
3.2 BQS Outliers
The eleven BQS SSQs lie in an atypical region of thecorrelation. This apparent anomaly is also presentin the radio–L[O III] plot (Rawlings 1994), so is un-likely to be (for instance) due to photometric er-rors in the BQS. There are two points to madeabout this feature, which we will support in thenext section with numerical simulations. First, thereis a selection effect which necessarily over-populatesthis region: the BQS, which covers half the sky, se-lects the rare quasars with the most extreme MB
at any z, favouring the high MB side of a radio–optical relation with large intrinsic scatter. In con-trast, the deeper MAQS is limited to ∼ 1 stera-
apparent tightening of the Qphot vs. Qbulk correlation seenat high-z in 3CRR (e.g. Rawlings & Saunders 1991).
Figure 5: The radio–optical relation for SSQs (symbols as Figure 1).
dian. Second, the BQS SSQs are typically ∼ 10times brighter in the radio than the faintest endof the radio LF, L408 ∼ 1024.5 W Hz−1 sr−1, wherethe comoving space density is an order-of-magnitudehigher (Dunlop & Peacock 1990). The lack of BQSSSQs with fainter radio luminosities is difficult toexplain without either a radio-optical correlation,or a strongly luminosity-dependent quasar fraction(Lawrence 1991; Jackson & Rawlings 1997; Serjeant et
al. in preparation). For the quasar fraction to explainthe deficit of low-L408 BQS SSQs, the SSQ luminosityfunction would have to be non-monotonic (i.e. num-ber density must not be a strictly decreasing functionof luminosity). There is no evidence for this within theMAQS (Serjeant et al. in preparation), though it ispossible that fainter radio samples may find such a lu-minosity cut-off (e.g. Willott et al. 1997). This leavesthe radio–optical correlation as the more plausible ex-planation. Furthermore, the fact that four of the BQSSSQs are also 3C radio sources (although only twomeet the more stringent selection criteria of 3CRR),and are thus among the most luminous radio sourcesin the sky, is clear evidence for the radio–optical cor-relation. Nevertheless, in section 3.3 we also show thatthe the top left hand corner of Figure 5 may be less
well sampled for SSQs than other regions, so the evi-dence of a radio-optical correlation in this part of theradio-optical plane is less compelling; a much moreconvincing demonstration is found to follow from thelack of quasars in the bottom right hand corner.
3.3 Simulated radio–optical relations
As the previous discussion of the BQS outliers illus-trates, the passage of flux limits across the radio-optical plane make a qualitative understanding of theselection effects on the radio–optical relation ratherdifficult. To clarify the situation we made numeri-cal simulations of the data. Note that a quantitativecomparison of the radio and optical properties of oursamples would involve estimating the bivariate radio-optical luminosity function, which will be discussedelsewhere (Serjeant et al., in preparation); here werestrict ourselves to reproducing only the gross prop-erties of our samples.
In Figure 6 we show the results of Monte-Carlosimulations of the MAQS, 3CRR and BQS samples.Points were sampled randomly from the three dimen-sional probability distribution defined by the bivariate(radio–optical) luminosity function. In Figure 6a, we
Figure 6a: Simulated data without a radio-optical correla-tion.
assume there is no radio–optical correlation, i.e. thebivariate LF is
Φ(L408, MB, z) ∝ ǫ(MB, z)β(L408, z) (1)
with the optical LF, ǫ, taken from Boyle et al. (1990),and the radio LF, β, taken from the “LDE” model ofDunlop & Peacock (1990). We renormalised the Boyleet al. LF to unity, i.e.
∫ǫ(MB, z)dMB = 1 (2)
Figure 6b: Simulated data with a radio-optical correlation
at all z; thus we are only using the shape of the opticalLF, and the cosmic evolution is determined by theradio LF alone. Note that the MAQS SSQs in the(MB, z) plane of (a) follow the bJ plate limit closely,whereas the 3CRR quasars have no optical flux limit.For simulation (b) (Figure 6b) we use
Φ(L408, MB, z) ∝ β(L408, z)γ(L408, MB) (3)
with the radio–optical correlation γ modelled asGaussian scatter of 1.5 optical magnitudes about theplotted full line, and again adopting the LDE LF βfrom Dunlop & Peacock (1990). In both simulationswe used α = 0.85 to convert from 2.7 GHz to 408MHz.
In both (a) and (b) we sampled 35 points from3CRR, 118 from MAQS and 11 from BQS, integrating
Φ throughout 24.5 < log10(P/(W Hz−1 sr−1)) < 29,−20 > MB > −29. Simulation (b) reproduces ourdata qualitatively, including the BQS outliers; thequantitative differences (particularly in the BQS zdistributions) are easily explained if the SSQ opticalLF
∫Φ dL408 evolves differently to the total quasar
LF (e.g. La Franca et al. 1994), or if the SSQ ra-dio LF differs from that of radio galaxies (Serjeant et
al. in preparation), or perhaps to unknown selectioneffects in the BQS itself (Goldschmidt et al. 1992).The qualitative agreement was found to be robust tothe assumed slope or dispersion in the radio–opticalcorrelation. Simulation (a), however, is grossly incon-sistent with our data: a general problem for mod-els without a radio–optical correlation is their inabil-ity to explain why SSQs from a bright radio sample(e.g. 3C) are among the brightest optically, and (moremarginally) vice versa (e.g. why the BQS containshigh-L408 SSQs).
An interesting feature of simulation (b) is the lackof evidence for a radio-optical correlation at high red-shifts, despite the fact that a correlation is present.In the simulation this is due to the lack of dynamicrange in radio power at high redshifts; the correlationhas a wide intrinsic dispersion and is not detectablewithout a wide range in radio power to compensate. Asimilar behaviour is seen in our 3CRR/MAQS/MQScombined sample at high redshift (Figure 2), againwhere the radio power dynamic range is least (Fig-ure 1). Also, in both our data and in simulation (b),the 3CRR quasars lie slightly on the radio-bright sideof the radio-opical correlation. This may be analagousto the BQS outliers (section 3.2): 3CRR selects theradio-brightest quasars at any z, favouring the highL408 side of a broad radio-optical correlation.
Finally, a very clear (perhaps the clearest) visualdemonstration of our correlation is obtained by com-paring Figure 5 with the MB − L408 plane of simu-lation (a) (Figure 6a, lower panel). The 3CRR SSQshave no optical flux limit, so in the absence of a corre-lation there is nothing to prevent them having opticalluminosities at the faintest end of the SSQ optical LF.This is clearly the case in simulation (a). However, inour data (and in simulation (b)) we find the 3CRRSSQs have optical luminosities lying on roughly thesame radio-optical relation as MAQS/MQS. This isobviously difficult to account for unless the radio andoptical luminosities are related in some way.
In other words, the lack of SSQs in the bottomright hand corner of Figure 5 is real; this area hasbeen well-sampled for SSQs. This is clearly highly sig-nificant, which we can quantify as follows. In 3CRRthere are 31 SSQs in the range 1026 < L408 < 1027.5
W Hz−1 sr−1, so if we assign absolute magnitudesuniformly over the range plotted in Figure 5 we findthe probability that the area is empty to be about10−4. This probability would be much smaller still ifwe were to use a more realistic optical LF.
In contrast, the top left corner may arguably
be less well sampled. The deficit of optically brightMAQS quasars could easily be attributed to the op-tical luminosity function of SSQs with faint radio lu-minosities. This leaves the BQS as the only surveywhich might adequately sample this region. The BQSdoes indeed lack the radio-faint, optically-bright SSQswhich would fill this region, in agreement with whatwe might expect from the radio-optical correlation.However, the lack of these BQS quasars could per-haps be related to the incompleteness worries in theBQS (Goldschmidt et al. 1992). Several authors (e.g.,La Franca et al. 1994) have also noted that the totalquasar population appears to have a much higher frac-tion of SSQs than at higher redshifts. One suggestedexplanation of this, which may also help explain theapparent incompleteness in the BQS, is that at suchlow redshifts the host galaxies may contribute non-negligibly to the total magnitudes. As a result, theBQS (by eye) stellar selection may bias the samplewith respect to host galaxy properties at these red-shifts in a complicated manner.
3.4 Comparison with previous results
At this point is is worth reviewing previous studiesin the light of our correlation, and contrasting themuch more problematic selection effects in previoussamples. A common problem is the inability to dis-tinguish evolution from luminosity dependence. Forexample, in section 1 we discussed the apparent cor-relation in 3CRR between bulk kinetic jet power Qbulk
and the photoionising radiation power output Qphot
as estimated from the narrow line luminosity LNLR
(Rawlings & Saunders 1991). Also, Qbulk and Qphot
have similar orders of magnitude in 3CRR, again sug-gesting a link. However, we will show that the selec-tion criteria of 3CRR could cause both this and theQbulk-Qphot correlation, if 3CRR it taken in isolation.
The Qbulk depends strongly on the total radio lu-minosity, which in turn correlates strongly with red-shift in 3C. This secondary correlation could ulti-mately lead to spurious relationships within 3C. Forexample, suppose there is no intrinsic Qbulk–Qphot
correlation (implying that the correlation in 3CRR isdue to some selection effect). Also, make the reason-able assumption that the optical luminosity functionof radio-loud quasars evolves roughly as strongly asits radio-quiet counterpart. Then the narrow-line lu-minosity LNLR will correlate strongly with redshift in3CRR, because LNLR is evolving, and so the Qbulk–Qphot correlation in 3CRR would be due entirely totheir independent evolution. The apparent similar-ity of the orders of magnitude of Qbulk and Qphot
throughout 3CRR may also be due to this indepen-dent evolution, and their apparent similarity wouldnot be preserved in samples of fainter radio luminos-ity. A similar critique can be made of the radio-quietQbulk-Qphot relation, though in this case the samples
are optically selected rather than radio flux limited.⋆
However, the presence of a radio-optical correlationin our samples demonstrates for the first time thatthe Rawlings & Saunders (1991) result is not due tothese selection effects, confirming their interpretation,and allows us to predict that the Qbulk − Qphot cor-relation will be preserved in complete samples withfainter limiting radio flux density.
If we assume our observed dispersion in the radio-optical correlation is close to the intrinsic value, thenwe can also resolve some of the previous contradictoryclaims on the existence of the correlation.
Peacock, Miller & Longair (1986) ruled out thenull hypothesis that the apparently bimodal distri-bution of radio-loud and radio-quiet quasars is infact due to a universal quasar radio-optical correla-tion (section 1). Our data can also immediately ruleout the alternative model suggested by Miller, Pea-cock & Mead (1990), that all radio-loud quasars havea minimum optical luminosity of MB ≃ −23. Also,on examining their sample selection it becomes clearwhy this study failed to detect the correlation. TheMiller et al. RLQ sample was wisely restricted to anarrow redshift range, which counters any differentialevolution, but unfortunately was limited to only sixSSQs which spanned less than an order of magnitudein both radio and optical luminosities. Given the verybroad dispersion apparent in our correlation, it hardlysurprising that they did not detect it. Interestingly,although the Peacock et al. study did not explicitlyaddress the SSQ radio-optical correlation, they notedthat several of the quasars in their deeper Parkes sub-samples have lower optical luminosities than those inbrighter Parkes subsamples. The samples were how-ever based in part on compilations from the Veron &Veron (1983) catalogue, and no attempt was made toseparate SSQs from flat-spectrum quasars.
Other previous inhomogeneous compilations alsogave evidence for an SSQ radio-optical correlation.Neff, Hutchings & Gower (1989) found a clear radio-optical correlation distinct from evolution effects,though their sample was selected from the literatureto fill the 2.7GHz radio luminosity, redshift planeas evenly as possible in 1 < z < 2. While this avoidsthe tendency towards fainter fluxes inherent in flux-limited samples, it is not clear if this method intro-duces selection effects of its own since the parentsample is clearly inhomogenous. Browne & Murphy1987 also found a clear radio-optical correlation inlobe-dominated quasars, though their sample was theradio-selected quasars in the Veron & Veron (1983)catalogue with published Einstein X-ray observations,which the authors emphasised “is a very heterogenoussample with all sorts of unknown selection effects.”
On the other hand, Browne & Wright (1985) had
⋆ Note though that the differing Qbulk/Qphot ratios inradio-quiet and radio-loud quasars is robust.
the benefit of complete radio flux limited samples, butlacked complete spectroscopic redshifts. The radio-optical flux-flux correlation present in their Figure 1could then easily be explained by differing redshiftdistributions, instead of an intrinsic radio-optical cor-relation. For instance, one might reasonably expectthat fainter samples extend to higher redshifts, sowould have correspondingly fainter optical identifica-tions. Such an interpretation is ruled out explicitly insection 3.1 above.
Miller et al. (1993) presented a study of the radioand optical properties of z < 0.5 BQS quasars, thoughthe number SSQs was again too small to detect ourcorrelation (see also Figure 5). However, we have alsoalready noted that the completeness of the BQS hasbeen questioned (Goldschmidt et al. 1992).
In summary, the only previous SSQ samples withredshift-independent evidence for a radio-optical cor-relation were inhomogeneous compilations, so proneto unquantified (and probably unquantifiable) selec-tion effects. The complete SSQ samples, on the otherhand, were either too small or too limited in radioor optical dynamic range to detect the correlation wehave found.
4 DISCUSSION
4.1 A link between accretion and the
fuelling of relativistic jets
We argue here that the SSQ radio-optical correlation(as predicted e.g. in the jet formation models of Fer-reira & Pelletier 1995) suggests a close link betweenthe formation of the jets and accretion onto the cen-tral black hole. Our discussion is similar to Rawlings& Saunders (1991), although now with the benefit ofmore direct evidence for a link between accretion andthe fuelling of relativistic jets and without the selec-tion effect ambiguities.
The narrow range in equivalent widths of broademission features (e.g. Miller et al. 1993, Francis et
al. 1993) and continuum variability studies imply thatthe bulk of the continuum is produced on sub-parsecscales, and is most naturally linked to accretion ontoa black hole (e.g. Begelman, Blandford & Rees 1984).Therefore, the SSQ radio-optical link appears to ariseon sub-parsec scales with the optical light producedby accretion, and with the radio luminosity derivedfrom a centrally-formed jet. This confirms the viewthat the dominant influence on L408 is Q, and notthe large scale radio source environment (Rawlings1993).
It is of course possible that the accretion-jet linkis not directly causal, since both processes could sharea close link with a third parameter. Any correlationof the type shown in Figure 5 is often dismissed as a“brighter objects are brighter” effect. However, thisscenario necessarily requires the existence of somelinks to create the scalings, and in the case of the
radio-optical correlation where both luminosities aregenerated by processes within the central parsec, sucha link is very likely to be close.
A radio-optical correlation could also be obtainedif both MB and L408 for individual quasars evolveseparately, but in the same sense, with time. How-ever, if the radio lobes of SSQs are short lived (i.e.lifetimes ≪ H−1
0 , e.g. Alexander & Leahy 1987), re-quiring multiple generations of SSQs, then a conspir-acy would have to be preserved from z ∼ 3 to z ∼ 0.4despite strongly changing physical environments (e.g.Yee & Green 1987; Ellingson & Green 1991; Haehnelt& Rees 1993).
The radio-optical correlation supports models inwhich SSQ jets are fuelled primarily by accretion onto a black hole (e.g. Begelman, Blandford & Rees1984), perhaps via magnetically driven winds (e.g.Spruit 1996; Ferreira & Pelletier 1995). In general,if any other parameters dominate the jet mechanism(such as disk angular momentum, disk structure, orrelated magnetic fields), then our correlation impliesthat they must also control or be controlled by theaccretion rate. In particular, this allows us to excludemany classes of models in which SSQ jets are fuelledprimarily by black hole spin energy and not accre-tion energy (Rawlings & Saunders 1990; Begelman,Blandford & Rees 1984). Only if the black hole spinregulates the accretion flow on (possibly) up to kpcscales can such models be sustained; Begelman (1985)presents a model in which he argues that local regu-lation of the accretion rate may be present.
It is also worth remembering that radio-quietquasars follow their own, but clearly different,radio–optical correlation (although it has not yetbeen shown to be redshift independent), having farlower jet powers than comparable radio-loud quasars(Miller et al. 1993; Lonsdale et al. 1995). Therefore, atleast one of the above parameters differ in radio-quietand radio-loud quasars.
4.2 Scatter in the radio–optical correlation
There are many possible physical interpretations forthe broad scatter in our correlation. We will list someof the more obvious candidates here.
First, radio luminosity gives an imprecise mea-sure of the bulk power in the jets (Qbulk) emanatingfrom the central engine since it also depends on thegaseous environment of the radio source (Rawlings &Saunders 1991; Ellingson et al. 1991); a plot of Q vs.
MB may give considerably less scatter than Figure 5.This would be consistent with the smaller dispersionin the Q-LNLR relation (Rawlings & Saunders 1991)for 3CRR radio sources. Unfortunately, at present welack the radio data necessary to make the conver-sion from L408 to Q for the MRC and 3CRR SSQs;the compact, steep-spectrum minority may move themost in the conversion to Q because their strong con-finement and/or young ages make the L408–Q con-
version extremely important. These quasars may alsoobey a different radio–optical relation if they have sig-nificantly fainter or redder continuua than the mainpopulation (e.g. Baker & Hunstead 1995).
Second, reddening of the optical quasar light isvery likely contributing to the scatter (Baker & Hun-stead 1995; Baker 1996). It may be possible to reducethis cause of scatter by deducing reddening from theoptical spectra, but since this typically needs higher-quality spectra and/or near-infrared photometry forour MAQS sample we have not yet been able to in-vestigate this.
Third, optical variability may also increase theobserved scatter and may also depend on luminos-ity. However, the variations are expected to be smallfor SSQs, about 0.2 mag on average on timescales ofyears (Hook et al. 1994). Variability scatter would bereduced by using median magnitudes over a longertimescale. The intrinsic spread in spectral energy dis-tributions over the observed optical waveband mustalso contribute to the scatter (e.g. Elvis 1994).
Fourth, the sampling of the radio–optical plane isstrongly non-uniform, as discussed above, which maybe responsible for some of the apparent dispersion.
Last, the dispersion could reflect additional de-pendences on secondary parameters. If this final sug-gestion is correct, then other observable quantities inSSQs may correlate with their deviation from the bestfit radio–optical correlation. Such a discovery mayprovide far stronger observational constraints on theformation of the jets.
4.3 The optical properties of radio galaxies
The radio–optical correlation we have established forSSQs has important implications for the orientation-based unified schemes for active galaxies (e.g. Barthel1989; Antonucci 1993). The radio–optical relationsuggests that amongst objects with a naked quasarnucleus those with the most luminous radio sourcesare necessarily also highly luminous in the optical. Ifthe compact quasar nucleus is hidden from our di-rect view, as is now known to be the case in someradio galaxies (e.g. Dey & Spinrad 1996), then in thecontext of the orientation-based unified schemes theradio–optical relation predicts that a high narrow-lineluminosity LNLR is inescapable. The radio–optical re-lation is then an obvious candidate for the causeof the relation between narrow-line luminosity LNLR
and L408 for radio sources (Baum & Heckman 1989;Rawlings & Saunders 1991; McCarthy 1993; Rawlings1993; Jackson & Rawlings 1997).
The SSQ radio–optical relation may also under-pin many anomalous properties of the most luminoushigh-z radio galaxies. For example, most models forthe alignment of the radio and optical structures ofhigh-z radio galaxies (e.g. McCarthy 1993) requirethat roughly constant fractions of Qbulk and MB areused to supply the different aligned components. If
we adopt the view that Q and MB are closely linked,then as L408 drops so must the level of aligned light,whether it is supplied by Qbulk from the jets (e.g.Lacy & Rawlings 1994) or by MB from the quasar(e.g. by scattering; Tadhunter et al. 1992). The lackof alignments and the low fraction of scattered opti-cal/UV light in less-luminous radio sources (Dunlop& Peacock 1993; Cimatti & di Serego Alighieri 1995),independent of redshift, support this prediction.
5 CONCLUSIONS
Using complete samples of SSQs, we have shown thatthe radio and optical luminosities of SSQs correlatewith ≫ 99% significance, independent of survey se-lection and cosmic epoch. Such a correlation wouldnaturally explain the observations of aligned opticalemission in only the most luminous radio galaxies.Also, we infer that the radio jets of SSQs are regu-lated by at least one parameter which is shared withthe production of the optical continuum; in the ac-cepted standard model for active galactic nuclei thisimplies a link between accretion and the fuelling ofthe relativistic jets.
This is not the first observational suggestion ofsuch a link, nor the first claim of a radio-optical corre-lation. However, this is the first evidence that neitherthe correlation nor the apparent radio-optical linksare caused by selection effects, particularly those in-herent in single flux-limited samples such as 3CR.
ACKNOWLEDGEMENTS
We thank the staff at the Anglo-Australian, WilliamHerschel and Nordic Optical Telescopes for technicalsupport. The WHT is operated on the island of LaPalma by the Royal Greenwich Observatory in theSpanish Observatorio del Roque de los Muchachos ofthe Instituto de Astrofisica de Canarias. SS thanksPPARC for a studentship, and for financial supportunder grant GR/K98728. We would also like to thankJohn Miller and R. Saunders for stimulating discus-sions, and the anonymous referee for careful readingof the manuscript and several helpful suggestions.
REFERENCES
Abramowicz M.A., Miller J.C. & Stuchlıc Z., 1993, Phys.
Rev. D, 47, 1440Alexander P. & Leahy J.P., 1987, Mon. Not. R. astr. Soc.
225, 1
Antonucci R., 1993, Ann. Rev. Astron. Astrophys., 31, 473Antonucci R., Hurt T. & Kinney A., 1994, Nature, 371,
313Axon D.J., Unger S.W., Pedlar A., Meurs E.J.A., Whittle,
Ogle, P.M., Cohen, M.H., Miller, J.S., Tran, H.D., Fosbury,R.A.E., Goodrich, R.W., 1997, preprint, to appear inAstrophys. J. Lett. (http://xxx.soton.ac.uk/abs/astro-ph/9703153)
Peacock J.A., Miller L. & Longair M.S., 1986, Mon Not.
R. astr. Soc., 218, 265Rawlings S., 1993, In: Extragalactic Radio Sources – From
Beams to Jets, p.332, eds Roland J., Sol H. & PelletierG.), Cambridge University Press, Cambridge, England
Rawlings S. 1994, In: The First Stromlo Symposium: The
Physics Of Active Galaxies, A.S.P. Conf. Ser. Vol. 54Rawlings S., Lacy M., Sivia D.S. & Eales S.A., 1995, Mon.
Not. R. astr. Soc. 274, 428Rawlings S. & Saunders R., 1991 Nature 349, 138Schmidt M. & Green R., 1983, Astrophys. J., 269, 352Serjeant S., 1996, D.Phil. thesis, University of OxfordSpruit H.C., 1996, preprint, to appear in Physical Pro-
cesses in Binary Stars, R.A.M.J. Wijers, M.B. Davies,C.A. Tout, eds., Kluwer Dordrecht (NATO ASI series)
Tadhunter, C.N., Scarrott, S.M., Draper, P. & Rolph C.,1992, Mon. Not. R. astr. Soc., 256, 53
